CN115954099B - Cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters - Google Patents
Cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters Download PDFInfo
- Publication number
- CN115954099B CN115954099B CN202211448858.5A CN202211448858A CN115954099B CN 115954099 B CN115954099 B CN 115954099B CN 202211448858 A CN202211448858 A CN 202211448858A CN 115954099 B CN115954099 B CN 115954099B
- Authority
- CN
- China
- Prior art keywords
- gait
- model
- features
- data
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention belongs to the technical field of cerebral apoplexy diagnosis, and particularly discloses a cerebral apoplexy associated quantitative evaluation method based on multi-mode gait parameters, which comprises the following steps: collecting multidimensional gait parameter data of a subject: collecting multidimensional gait parameter data of a tested person by a gait analysis system, and analyzing and detecting to obtain time parameter characteristics and space parameter characteristics of the tested person; and (3) data normalization processing: the data is normalized by adopting a maximum and minimum method, and the purpose of the normalization processing is to map all the features into the range of (0, 1) so as to ensure that the images of the mechanical learning models after feature pairs with different dimensions and different orders of magnitude are the same; according to the invention, through all existing gait characteristics, proper gait characteristics are selected, the stroke patient is calibrated by the NHISS scale of a clinician, and a model which can be used for judging whether the stroke patient is and evaluating the stroke severity in a grading way is finally trained through a machine learning algorithm.
Description
Technical Field
The invention relates to the technical field of cerebral apoplexy diagnosis, in particular to a cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters.
Background
Traditionally, in determining whether a person has a stroke or classifying the severity of a stroke patient, experienced doctors often score case group subjects according to the national institutes of health stroke scale (NHISS) to assess the severity of the stroke.
In the previous study Kong Youqi, the gait analysis and detection were performed on the stroke patient and the healthy control group by using the lid force walking state analysis system, and it was found that the affected side of the stroke patient was compared with the control group, the stride of the case group was shortened, the stride frequency and the pace were decreased, and the swing phase time, the support phase time and the bipedal support time were prolonged (P < 0.01).
The current clinical gait analysis is mostly used for revealing that the gaits of the cerebral apoplexy patients have obvious differences in many aspects, and the rehabilitation quality effect of the cerebral apoplexy patients is effectively evaluated by determining the gait characteristics (gait parameters) of the cerebral apoplexy patients. The initial gait analysis method is a method of subjective observation by using experience of a clinician, or video observation, or scoring by a scale. The method has the advantages of simplicity, low cost, large qualitative and subjective performance, and poor accuracy and mainly depends on the experience of doctors. In addition, there are a small number of intelligent studies on the analysis of the gait of a patient suffering from cerebral apoplexy, such as discrimination of whether it is cerebral apoplexy based on symmetry, regularity and stability of gait as characteristics, and classification and evaluation studies on the gait of a patient suffering from cerebral apoplexy, but there are few cases based on more comprehensive gait space-time characteristics (parameters).
Disclosure of Invention
The invention aims to provide a cerebral apoplexy associated quantitative evaluation method based on multi-mode gait parameters, which aims to solve the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions: a cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters comprises the following steps:
s1, collecting multidimensional gait parameter data of a tested person: collecting multidimensional gait parameter data of a tested person by a gait analysis system, and analyzing and detecting to obtain time parameter characteristics and space parameter characteristics of the tested person;
s2, data normalization processing: the data is normalized by adopting a maximum and minimum method, and the purpose of the normalization processing is to map all the features into the range of (0, 1) so as to ensure that the images of the mechanical learning models after feature pairs with different dimensions and different orders of magnitude are the same;
s3, scoring and marking by an NIHSS scale: the doctor adopts an NIHSS scale for clinically evaluating the severity of the cerebral apoplexy patient to score the tested person and marks corresponding data;
s4, selecting proper characteristics: selecting proper characteristics according to the discrimination model and the grading model;
s5, selecting a proper machine learning classification model; the machine learning classification model comprises a discrimination model and a classification model, and the steps of the discrimination model and the classification model are the same;
s6, model training: model training comprises the specific steps of distinguishing models and grading models, wherein the steps are the same: model training and parameter tuning are carried out according to the multidimensional gait data after the normalization to be tested, labeling results of doctors under different tasks, proper characteristics and algorithms constructed by the model; and selecting tuning parameters by adopting a cross-validation method. Finally we can get the trained model A 1 ’、A 2 ’、A 3 ’、…、A j ' and parameters after tuning of each modelWherein k is model A j Is used for the number of adjustable parameters.
Preferably, the formula of the data normalization process is:
wherein x is the time parameter feature and the space parameter feature of the testee, and represents the value of the initial data, x' is the multidimensional gait data after normalization processing, represents the scaled value, and x max X is the maximum value in the time parameter characteristic and the space parameter characteristic of the tested person min Is the minimum value of the time parameter characteristic and the space parameter characteristic of the testee.
Preferably, the NIHSS scale consists of 11 evaluation indexes, wherein the score range of each evaluation index is 0-4 points, and 0 points are obtained to indicate that the evaluation index is normal in function, the higher the score is, the worse the function is, total NHISS score is generated after all the evaluation of the 11 evaluation indexes is finished, and a doctor marks whether a tested person suffers from cerebral apoplexy or not for judging tasks; for the grading task, the doctor marks the severity of the stroke being tested.
Preferably, the steps of selecting the proper features according to the discrimination model and the grading model are the same, and the method comprises the following steps:
1) For the normalized multidimensional gait data of each tested person, the single characteristics and the labeling results of doctors in different tasks are randomly grouped and input into M 1 Different machine learning models record and arrange the learning results to obtain M 1 Sorting the importance degrees of the different features, and calculating the average value of the importance degree sorting of each feature in all models to obtain the comprehensive sorting of the importance degree of each feature;
2) N features with top rank are selected from the comprehensive ranking of the importance degree of each feature, and M features are used respectively through a cross verification method 2 The N features are learned by different machine learning models, and the M is obtained 2 Model accuracy vectors using these N features under different models, denoted as M 2 N；
3) I different N (n=n) are selected 1 ，N 2 ，N 3 ，…，N i I is smaller than the maximum number of features), repeating the step 2) to obtain i M 2 N i Selecting M with best overall effect 2 N i Wherein N is i I.e. the number of selected features, the top N in the comprehensive ordering of the feature importance degree i The individual features are selected as appropriate.
Preferably, the specific steps for selecting an appropriate machine-learned classification model are as follows: classification algorithm for selecting models by cross-validation, in particular selecting common M 3 The method comprises the steps of calculating performance indexes of the classification algorithms respectively, performing cross validation calculation for 5 times by each algorithm, calculating prediction accuracy of 5 times, obtaining the prediction accuracy of the algorithm, taking an average value of 5 times, performing 30 times on the operation, and finding out the classification algorithm with high peak value; under the condition of respectively comparing the algorithms, the obtained corresponding accuracy peak values are compared, and finally the algorithm A with higher accuracy peak value is selected 1 、A 2 、A 3 、…、A j Where j is the total number of models, which is the algorithm for our model construction.
Compared with the prior art, the invention has the beneficial effects that: aiming at the previously proposed step detection algorithm and gait characteristic (parameter) calculation method, the invention provides a stroke association quantitative evaluation method based on multidimensional gait parameters. The method is based on a machine learning method, a model is trained, and gait analysis is further carried out. Namely, through all existing gait characteristics (gait parameters), proper gait characteristics (gait parameters) are selected, a judgment model (judgment model) for judging whether a cerebral apoplexy patient is or not and a model (grading model) for grading evaluation of cerebral apoplexy severity are finally trained through the calibration of a clinical doctor's NHISS scale on a cerebral apoplexy patient and a machine learning algorithm.
The brain stroke association quantitative evaluation method based on the multidimensional gait parameters has higher accuracy and is expected to be used for early warning, rehabilitation monitoring and evaluation of recurrence after brain stroke in clinic so as to assist doctors.
Drawings
FIG. 1 is a process and flow diagram showing the selection of suitable features of the present invention;
FIG. 2 is a visual ranking of feature importance levels selected by random forest, adaBoost, gradientBoosting, decisionTree algorithm in the discrimination model of the present invention, wherein the abscissa is the feature relative importance level;
FIG. 3 is a visual ordering of feature importance selected by the random forest, adaBoost, gradientBoosting, decisionTree algorithm in the hierarchical model of the present invention, wherein the abscissa is the relative importance of features.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
226 cases of data (including normal healthy people and cerebral apoplexy patients) are collected through neurology department of Beijing Chaoyang hospital (western hospital), and experiments are respectively carried out on the models constructed herein to verify the effectiveness of the models. The experiment is divided into two parts, the first part (hereinafter referred to as experiment one) is a verification and discrimination model, namely discrimination between normal healthy people and cerebral apoplexy patients, and the second part (hereinafter referred to as experiment two) is a verification and classification model, namely discrimination on cerebral apoplexy patients.
S1, collecting multidimensional gait parameter data of a tested person
The experimental data are 226 cases of data collected at Beijing Kogyo Hospital affiliated to the university of capital medical science, the experimental process and notice are explained to the subjects before the experiment starts, the details of the subjects participating in the experiment are shown in the following table 1 (table 1 is experimental data information), and the data are taken as a data set of the experiment I of the patent;
gait characteristics in the patent adopt a high-precision step detection and extraction algorithm based on wearable heterogeneous sensor signal fusion and a clinically-reliable gait parameter calculation method based on multi-strategy information fusion, 32 gait characteristic parameters are calculated in total, and the gait characteristic parameters are classified as follows according to parameter categories:
time parameter characteristics (8): left swing time (left swing phase), right swing time (right swing phase), left support time (left stance phase), right support time (right stance phase), left single foot time (left time), right single foot time (right time), left double foot time (left double time), right double foot time (right double time).
Spatial parameter characterization (24): left stride (left stride), right stride (right stride), left Bu Pin (left step frequency), right stride frequency (right step frequency), left stride speed (left step speed), right stride speed (right step speed), left foot deflection angle (left foot rotation), right foot deflection angle (right foot rotation), left ground clearance angle (left off ground angle), right ground clearance angle (right off ground angle), left ground clearance angle (left ground angle), right ground clearance angle (right ground angle), left sagittal minimum angle (left sagittal min angle), right sagittal minimum angle (right sagittal min angle), left sagittal maximum angle (left sagittal max angle), right sagittal maximum angle (right sagittal max angle), left coronal minimum angle (left coronal min angle), right coronal minimum angle (right coronal min angle), left coronal maximum angle (left coronal max angle), right coronal maximum angle (right coronal max angle), left transverse minimum angle (left cross min angle), right transverse minimum angle (right cross min angle), left transverse maximum angle (left cross max angle), right transverse maximum angle (right cross max angle) (the calculated angles are all based on calculated inertial sensor parameters.
S2, data normalization processing
And (3) converting the data acquired in the step (S1) by using a normalization formula respectively for each feature.
S3, marking and labeling of NIHSS (network subscriber server) scale
In Table 1, 32 of 39 stroke patients were scored and calibrated by the clinician using the NIHSS scale, and the scores of the calibrated patients were distributed between 1-5 points, i.e., all non-severe stroke patients, specifically as shown in Table 2 below (Table 2 is the NIHSS score of the stroke patient); during testing, the patient is required to walk on a flat ground along a straight line, and the patient is required to walk for several times as much as possible according to the actual situation of the patient, namely, the walking times of different patients are different, and each walking process covers about 2-50 steps;
in the patent, parameters of each step of each patient in each walking are calculated to form a data set of the experiment II, and the total step number of 32 patients is 515; considering the symmetry of the samples, we classified those with NIHSS scores of 1 as lighter stroke patients and NIHSS scores of 2-5 as mild stroke patients. Specific experimental data of the second experiment are shown in table 3 (table 3 is a stroke patient weight judging experimental data set);
s4, selecting proper characteristics
Features are screened by using Random forest and AdaBoost, gradientBoosting, decisionTree algorithms, and N features which are most important are selected. The method has the thought that the machine learning model is directly used for selecting the characteristics, a prediction model is suggested according to each independent characteristic and response variable, the model is selected by cross verification, and the specific steps and the flow are shown in the figure 1;
firstly, learning single features in normalized multidimensional gait data through a random forest algorithm and a AdaBoost, gradientBoosting, decisionTree algorithm, sorting the importance degrees of the selected features, and obtaining comprehensive sorting of the importance degrees of the features after averaging.
Secondly, considering that the total 32 feature parameters exist in the data set, in order to avoid the problems caused by overfitting and underfilling caused by too much and too little feature selection, we respectively select the top 10 features, the top 20 features and the top 32 features (all features) from the ranking according to the importance, and respectively perform cross-validation by using a LogisticRegression, gaussianNB, decisionTree, randomForest, SVM algorithm to obtain the accuracy of the features. And then comparing the accuracy of 10 features, 20 features and 32 features (all features) under the condition of each algorithm respectively, and comprehensively judging the number of the features selected by the final classifier.
Discrimination model: according to the above feature selection algorithm, 10 features, 20 features and 32 features are first selected and cross-validated by 5 common algorithms, respectively, to obtain their corresponding accuracies as shown in table 4 (table 4 is a feature selection performance comparison table). From the table, when 10 features are selected, the classifier has better classifying effect and smaller calculated amount. Fig. 2 is a visual ordering of feature importance (feature relative importance on the abscissa) selected by the random forest, adaBoost, gradientBoosting, decisionTree algorithm. In this way, we finally select 10 features that constitute a feature vector: right step speed (right step speed), left step speed (left step speed), right ground clearance angle (right off ground angle), left ground clearance angle (left off ground angle), right sagittal minimum angle (right sagittal min angle), left sagittal minimum angle (left sagittal min angle), right stride (right stride), left stride (left stride), right sagittal maximum angle (right sagittal max angle), left sagittal maximum angle (left sagittal max angle);
| algorithm | 10 features | 20 features | 32 features |
| LogisticRegression | 90.42％ | 87.19％ | 82.09％ |
| GaussianNB | 88.46％ | 86.49％ | 87.19％ |
| DecisionTree | 84.09％ | 82.67％ | 78.76％ |
| RandomForest | 89.75％ | 88.46％ | 89.15％ |
| SVM | 89.77％ | 89.47％ | 88.55％ |
And (3) grading model: according to the feature selection algorithm, 10 features, 20 features and 32 features are selected first, and cross-validation is performed by 5 common algorithms, so that the corresponding accuracy rates are shown in the following table 5 (table 5 feature selection performance comparison table). From the table, when 10 features are selected, the classifier has better classifying effect and smaller calculated amount;
| algorithm | 10 features | 20 features | 32 features |
| LogisticRegression | 44.30％ | 43.50％ | 43.59％ |
| GaussianNB | 47.98％ | 47.51％ | 46.59％ |
| DecisionTree | 55.75％ | 54.61％ | 52.97％ |
| RandomForest | 67.10％ | 66.19％ | 62.03％ |
| SVM | 64.25％ | 63.68％ | 64.11％ |
Fig. 3 is a visual ordering of feature importance (feature relative importance on the abscissa) selected by the random forest, adaBoost, gradientBoosting, decisionTree algorithm. In this way, we finally select 10 features that constitute a feature vector: a healthy side grounding angle (healthy side ground angle), a healthy side grounding angle (affected side ground angle), a healthy side ground clearance angle (healthy side off ground angle), a healthy side ground clearance angle (affected side off gound angle), a healthy side cross section maximum angle (healthy side cross max angle), a healthy side cross section maximum angle (affected side cross max angle), a healthy side step size (healthy side stride), a healthy side bipedal time (affected double time), a healthy side sagittal plane maximum angle (healthy side sagittal max angle), and a healthy side sagittal plane maximum angle (affected side sagittal max angle).
S5, selecting a proper machine learning classification model
After determining the characteristics of the model training, a classification algorithm is selected. The classification algorithm of the model is selected by a cross-validation method, particularly 9 common classification algorithms are selected, and the performance indexes of the classification algorithms are calculated respectively. Each algorithm performs 5 times of cross validation calculation, calculates 5 times of prediction accuracy, and obtains the prediction accuracy (average value of 5 times) of the algorithm, and performs 30 times of operation to find out the classification algorithm with high peak value. And under the condition of respectively comparing the algorithms, the obtained corresponding accuracy peak values are compared, and finally, the algorithm with higher accuracy peak value is selected as the algorithm constructed by the model.
Discrimination model: according to the determination method selected by the classification algorithm, each algorithm and peak values are shown in the following table 6 (table 6 is an experimental performance comparison table for algorithm selection of a discrimination model):
through comparison, the following 3 algorithms are selected for classifier modeling: KNN, SVM and random forest.
And (3) grading model: according to the determination method selected by the classification algorithm, each algorithm and peak values are shown in the following table 7 (table 7 is an experimental performance comparison table for algorithm selection of the classification model):
| algorithm | Peak performance |
| KNN[106-108] | 63.97％ |
| Perceptron[109] | 57.25％ |
| AdaBoost | 66.88％ |
| Stochastic Gradient Descent[110] | 60.76％ |
| LogisticRegression | 43.93％ |
| SVM | 69.35％ |
| DecisionTree | 57.05％ |
| RandomForest | 67.32％ |
| GradicntBoosting | 60.76％ |
Through comparison, the following 3 algorithms are selected for classifier modeling: SVM, randomForest and AdaBoost.
S6, model training
The model training is to construct a classifier, namely, the process of model training and parameter tuning is carried out according to the data set, the characteristics and the algorithm selected before, and the output of the classifier is the constructed model. The patent adopts a 5-fold cross validation method to carry out parameter tuning and selection. The basic idea is to group training sets in the original data set, one part is used as the training set to train the model, and the other part is used as the test set to evaluate the model (i.e. the verification set).
Discrimination model: as shown in table 1, we selected 156 patients and normal persons as training sets to construct a discriminant model, and the obtained relevant parameters are: the parameter n_neighbors=6 of KNN, the svm parameter c=2.1, max_iter=21, kernel=poly, the random forest parameter n_estimators=9.
And (3) grading model: as shown in table 3, we selected a total of 515 steps for 32 patients, where the training set consisted of 457 steps for 25 patients, to construct a hierarchical model that gave the relevant parameters for three different algorithms: adaboost's parameter n_estimators=7, learning_rate=1.8, SVM parameter C=0.1, max_iter=7, kernel=Poly, random forest parameter n_estimators=52.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (4)
1. A cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters is characterized by comprising the following steps:
s1, collecting multidimensional gait parameter data of a tested person: collecting multidimensional gait parameter data of a tested person by a gait analysis system, and analyzing and detecting to obtain time parameter characteristics and space parameter characteristics of the tested person;
s2, data normalization processing: the data is normalized by adopting a maximum and minimum method, and the purpose of the normalization processing is to map all the features into the range of (0, 1) so as to ensure that the images of the mechanical learning models after feature pairs with different dimensions and different orders of magnitude are the same;
s3, scoring and marking by an NIHSS scale: the doctor adopts an NIHSS scale for clinically evaluating the severity of the cerebral apoplexy patient to score the tested person and marks corresponding data;
s4, selecting proper characteristics: selecting proper characteristics according to the discrimination model and the grading model, wherein the specific steps are the same, and the method comprises the following steps:
1) For the normalized multidimensional gait data of each tested person, the single characteristics and the labeling results of doctors in different tasks are randomly grouped and input into M 1 Different machine learning models record and arrange the learning results to obtain M 1 Sorting the importance degrees of the different features, and calculating the average value of the importance degree sorting of each feature in all models to obtain the comprehensive sorting of the importance degree of each feature;
2) N features with top rank are selected from the comprehensive ranking of the importance degree of each feature, and M features are used respectively through a cross verification method 2 The N features are learned by different machine learning models, and the M is obtained 2 Model accuracy vectors using these N features under different models, denoted as M 2 N；
Selecting i different N, n=n 1 ，N 2 ，N 3 ，…，N i I is less than the maximum number of features; repeating the step 2 to obtain i M 2 N i Selecting M with best overall effect 2 N i Wherein N is i I.e. the number of selected features, the top N in the comprehensive ordering of the feature importance degree i The individual features are selected as appropriate features;
s5, selecting a proper machine learning classification model; the machine learning classification model comprises a discrimination model and a classification model, and the steps of the discrimination model and the classification model are the same;
s6, model training: the model training comprises the following steps: model training and parameter tuning are carried out according to the normalized multidimensional gait data of the selected testee, the labeling results of doctors under different tasks, proper characteristics and the algorithm constructed by the model; the cross-validation method is adopted to select tuning parameters, and a trained model A1 can be obtained ， 、A2 ， 、A3 ， 、…、Aj ， 。
2. The method for quantitatively evaluating cerebral apoplexy related state based on multi-modal gait parameters according to claim 1, wherein the method is characterized in that: the formula of the data normalization processing is as follows:
wherein x is the time parameter feature and the space parameter feature of the testee, and represents the value of initial data, x ， To normalize the processed multidimensional gait data, scaled values, x max X is the maximum value in the time parameter characteristic and the space parameter characteristic of the tested person min Is the minimum value of the time parameter characteristic and the space parameter characteristic of the testee.
3. The method for quantitatively evaluating cerebral apoplexy related state based on multi-modal gait parameters according to claim 1, wherein the method is characterized in that: the NIHSS scale consists of 11 evaluation indexes, wherein the score range of each evaluation index is 0-4 points, the score of 0 points indicates that the evaluation index is normal in function, the score of higher points indicates that the function is worse, total NHISS score is generated after all evaluation of 11 evaluation indexes are finished, and a doctor marks whether a tested brain stroke exists or not for a judging task; for the grading task, the doctor marks the severity of the stroke being tested.
4. The method for quantitatively evaluating cerebral apoplexy related state based on multi-modal gait parameters according to claim 1, wherein the method is characterized in that: the specific steps of selecting a suitable machine learning classification model are as follows: classification algorithm for selecting models by cross-validation, in particular selecting common M 3 And (3) a classification algorithm, namely respectively calculating performance indexes of each classification algorithm, performing cross validation calculation for 5 times by each algorithm, calculating prediction accuracy for 5 times, obtaining the prediction accuracy of the algorithm, taking an average value for 5 times, then executing 30 times on the operation, respectively comparing the obtained peak values of the corresponding accuracy under the condition of each algorithm, and selecting the algorithm with the highest accuracy peak value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211448858.5A CN115954099B (en) | 2022-11-18 | 2022-11-18 | Cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211448858.5A CN115954099B (en) | 2022-11-18 | 2022-11-18 | Cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115954099A CN115954099A (en) | 2023-04-11 |
| CN115954099B true CN115954099B (en) | 2023-09-15 |
Family
ID=87288096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211448858.5A Active CN115954099B (en) | 2022-11-18 | 2022-11-18 | Cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115954099B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678762A (en) * | 2016-01-06 | 2016-06-15 | 四川大学 | System for monitoring treatment of ischemic cerebrovascular disorder on the basis of image information |
| CN113362944A (en) * | 2021-05-14 | 2021-09-07 | 西安交通大学 | Machine learning-based prediction of functional dyspepsia and acupuncture curative effect |
| CN114299996A (en) * | 2021-12-31 | 2022-04-08 | 南京邮电大学 | AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190259475A1 (en) * | 2018-02-20 | 2019-08-22 | SameDay Security, Inc. | Connected Kiosk for the Real-Time Assessment of Falls Risk |
| US20220293266A1 (en) * | 2021-03-09 | 2022-09-15 | The Hong Kong Polytechnic University | System and method for detection of impairment in cognitive function |
-
2022
- 2022-11-18 CN CN202211448858.5A patent/CN115954099B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678762A (en) * | 2016-01-06 | 2016-06-15 | 四川大学 | System for monitoring treatment of ischemic cerebrovascular disorder on the basis of image information |
| CN113362944A (en) * | 2021-05-14 | 2021-09-07 | 西安交通大学 | Machine learning-based prediction of functional dyspepsia and acupuncture curative effect |
| CN114299996A (en) * | 2021-12-31 | 2022-04-08 | 南京邮电大学 | AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease |
Non-Patent Citations (1)
| Title |
|---|
| 柔性足压反馈训练系统对卒中患者脑区功能重塑的纵向研究;叶强;刘莉;钱竞光;顾小东;高润;;南京医科大学学报(自然科学版)(04);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115954099A (en) | 2023-04-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108597601B (en) | Support vector machine-based chronic obstructive pulmonary disease diagnosis auxiliary system and method | |
| CN107753026B (en) | Intelligent shoe self-adaptive monitoring method for spinal leg health | |
| Khandoker et al. | A comparative study on approximate entropy measure and poincaré plot indexes of minimum foot clearance variability in the elderly during walking | |
| CN110367991B (en) | Old people falling risk assessment method | |
| CN1359656A (en) | Method and device for estimating neure damage degree in cerabral cortex | |
| US20200187842A1 (en) | Multiparameter method for quantifying balance | |
| Albert et al. | Using machine learning to predict perceived exertion during resistance training with wearable heart rate and movement sensors | |
| Chen et al. | Interpretable evaluation for the Brunnstrom recovery stage of the lower limb based on wearable sensors | |
| TWI688371B (en) | Intelligent device for atrial fibrillation signal pattern acquisition and auxiliary diagnosis | |
| Singh et al. | Detection of Parkinson's Disease using Machine Learning Algorithm | |
| CN117153379B (en) | Prediction device for thoracic outlet syndrome | |
| CN108447562B (en) | User exercise capacity assessment method and system | |
| CN115954099B (en) | Cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters | |
| CN117436002A (en) | Intelligent evaluation system for amblyopia type and severity based on eye movement characteristics | |
| Asaithambi et al. | Classification of respiratory abnormalities using adaptive neuro fuzzy inference system | |
| TWI796275B (en) | Method for disease risk assessment | |
| CN114171194B (en) | Quantitative assessment method, device, electronic device and medium for Parkinson multiple symptoms | |
| JP7454679B2 (en) | Binocular keratoconus evaluation system based on multimodal data | |
| CN115661101A (en) | Premature infant retinopathy detection system based on random sampling and deep learning | |
| Tao et al. | Activity selection to distinguish healthy people from parkinson’s disease patients using i-da | |
| Lim et al. | GAIT ANALYSIS AND CLASSIFICATION ON SUBJECTS WITH PARKINSON’ S DISEASE | |
| WO2021042372A1 (en) | Atrial fibrillation prediction model and prediction system thereof | |
| Rakshna et al. | Pre-Stroke Detection using K-Nearest Neighbour and Random Forest Algorithm | |
| Elwin et al. | Entropy Weighted and Kernalized Power K-Means Clustering Based Lesion Segmentation and Optimized Deep Learning for Diabetic Retinopathy Detection | |
| CN116746910B (en) | Gait monitoring method and device based on wearable equipment and wearable equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |